Perceptually adaptive joint deringing - deblocking filtering for scalable video coding

摘要

A new perceptually adaptive joint deringing - deblocking filtering method for wavelet based scalable video coding is proposed. The proposed approach is specially designed to deal with the artifacts inherent to video transmission over very low-bit-rate channels, wireless channel for example. A structure based on in-loop filtering is designed to decode scalable coded video considering the update step in motion compensated temporal filtering. Since ringing and blocking artifacts are visually disturbing, the characteristics of the human visual system (HVS) are considered in the underlying bilateral filtering. In the proposed approach, the filter strength is adjusted according to the perceptual distortion by integrating the HVS models of luminance masking, activity masking and temporal masking. Consequently, a different filtering strength is applied to areas with different perceptual sensitivity in terms of the HVS. A comparative evaluation using conventional loop filtering structure and bilateral filter was conducted. The experimental assessment demonstrates a superior performance of the proposed adaptive filtering, providing substantial and consistent better objective and subjective quality.